4 * Copyright (C) 1991, 1992 Linus Torvalds
7 #include <linux/string.h>
9 #include <linux/file.h>
10 #include <linux/fdtable.h>
11 #include <linux/fsnotify.h>
12 #include <linux/module.h>
13 #include <linux/tty.h>
14 #include <linux/namei.h>
15 #include <linux/backing-dev.h>
16 #include <linux/capability.h>
17 #include <linux/securebits.h>
18 #include <linux/security.h>
19 #include <linux/mount.h>
20 #include <linux/fcntl.h>
21 #include <linux/slab.h>
22 #include <asm/uaccess.h>
24 #include <linux/personality.h>
25 #include <linux/pagemap.h>
26 #include <linux/syscalls.h>
27 #include <linux/rcupdate.h>
28 #include <linux/audit.h>
29 #include <linux/falloc.h>
30 #include <linux/fs_struct.h>
31 #include <linux/ima.h>
32 #include <linux/dnotify.h>
36 #define CREATE_TRACE_POINTS
37 #include <trace/events/fs.h>
39 int do_truncate(struct dentry *dentry, loff_t length, unsigned int time_attrs,
43 struct iattr newattrs;
45 /* Not pretty: "inode->i_size" shouldn't really be signed. But it is. */
49 newattrs.ia_size = length;
50 newattrs.ia_valid = ATTR_SIZE | time_attrs;
52 newattrs.ia_file = filp;
53 newattrs.ia_valid |= ATTR_FILE;
56 /* Remove suid/sgid on truncate too */
57 ret = should_remove_suid(dentry);
59 newattrs.ia_valid |= ret | ATTR_FORCE;
61 mutex_lock(&dentry->d_inode->i_mutex);
62 ret = notify_change(dentry, &newattrs);
63 mutex_unlock(&dentry->d_inode->i_mutex);
67 static long do_sys_truncate(const char __user *pathname, loff_t length)
74 if (length < 0) /* sorry, but loff_t says... */
77 error = user_path(pathname, &path);
80 inode = path.dentry->d_inode;
82 /* For directories it's -EISDIR, for other non-regulars - -EINVAL */
84 if (S_ISDIR(inode->i_mode))
88 if (!S_ISREG(inode->i_mode))
91 error = mnt_want_write(path.mnt);
95 error = inode_permission(inode, MAY_WRITE);
97 goto mnt_drop_write_and_out;
100 if (IS_APPEND(inode))
101 goto mnt_drop_write_and_out;
103 error = get_write_access(inode);
105 goto mnt_drop_write_and_out;
108 * Make sure that there are no leases. get_write_access() protects
109 * against the truncate racing with a lease-granting setlease().
111 error = break_lease(inode, O_WRONLY);
113 goto put_write_and_out;
115 error = locks_verify_truncate(inode, NULL, length);
117 error = security_path_truncate(&path);
119 error = do_truncate(path.dentry, length, 0, NULL);
122 put_write_access(inode);
123 mnt_drop_write_and_out:
124 mnt_drop_write(path.mnt);
131 SYSCALL_DEFINE2(truncate, const char __user *, path, long, length)
133 return do_sys_truncate(path, length);
136 static long do_sys_ftruncate(unsigned int fd, loff_t length, int small)
138 struct inode * inode;
139 struct dentry *dentry;
151 /* explicitly opened as large or we are on 64-bit box */
152 if (file->f_flags & O_LARGEFILE)
155 dentry = file->f_path.dentry;
156 inode = dentry->d_inode;
158 if (!S_ISREG(inode->i_mode) || !(file->f_mode & FMODE_WRITE))
162 /* Cannot ftruncate over 2^31 bytes without large file support */
163 if (small && length > MAX_NON_LFS)
167 if (IS_APPEND(inode))
170 error = locks_verify_truncate(inode, file, length);
172 error = security_path_truncate(&file->f_path);
174 error = do_truncate(dentry, length, ATTR_MTIME|ATTR_CTIME, file);
181 SYSCALL_DEFINE2(ftruncate, unsigned int, fd, unsigned long, length)
183 long ret = do_sys_ftruncate(fd, length, 1);
184 /* avoid REGPARM breakage on x86: */
185 asmlinkage_protect(2, ret, fd, length);
189 /* LFS versions of truncate are only needed on 32 bit machines */
190 #if BITS_PER_LONG == 32
191 SYSCALL_DEFINE(truncate64)(const char __user * path, loff_t length)
193 return do_sys_truncate(path, length);
195 #ifdef CONFIG_HAVE_SYSCALL_WRAPPERS
196 asmlinkage long SyS_truncate64(long path, loff_t length)
198 return SYSC_truncate64((const char __user *) path, length);
200 SYSCALL_ALIAS(sys_truncate64, SyS_truncate64);
203 SYSCALL_DEFINE(ftruncate64)(unsigned int fd, loff_t length)
205 long ret = do_sys_ftruncate(fd, length, 0);
206 /* avoid REGPARM breakage on x86: */
207 asmlinkage_protect(2, ret, fd, length);
210 #ifdef CONFIG_HAVE_SYSCALL_WRAPPERS
211 asmlinkage long SyS_ftruncate64(long fd, loff_t length)
213 return SYSC_ftruncate64((unsigned int) fd, length);
215 SYSCALL_ALIAS(sys_ftruncate64, SyS_ftruncate64);
217 #endif /* BITS_PER_LONG == 32 */
220 int do_fallocate(struct file *file, int mode, loff_t offset, loff_t len)
222 struct inode *inode = file->f_path.dentry->d_inode;
225 if (offset < 0 || len <= 0)
228 /* Return error if mode is not supported */
229 if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE))
232 /* Punch hole must have keep size set */
233 if ((mode & FALLOC_FL_PUNCH_HOLE) &&
234 !(mode & FALLOC_FL_KEEP_SIZE))
237 if (!(file->f_mode & FMODE_WRITE))
240 /* It's not possible punch hole on append only file */
241 if (mode & FALLOC_FL_PUNCH_HOLE && IS_APPEND(inode))
244 if (IS_IMMUTABLE(inode))
248 * Revalidate the write permissions, in case security policy has
249 * changed since the files were opened.
251 ret = security_file_permission(file, MAY_WRITE);
255 if (S_ISFIFO(inode->i_mode))
259 * Let individual file system decide if it supports preallocation
260 * for directories or not.
262 if (!S_ISREG(inode->i_mode) && !S_ISDIR(inode->i_mode))
265 /* Check for wrap through zero too */
266 if (((offset + len) > inode->i_sb->s_maxbytes) || ((offset + len) < 0))
269 if (!file->f_op->fallocate)
272 return file->f_op->fallocate(file, mode, offset, len);
275 SYSCALL_DEFINE(fallocate)(int fd, int mode, loff_t offset, loff_t len)
282 error = do_fallocate(file, mode, offset, len);
289 #ifdef CONFIG_HAVE_SYSCALL_WRAPPERS
290 asmlinkage long SyS_fallocate(long fd, long mode, loff_t offset, loff_t len)
292 return SYSC_fallocate((int)fd, (int)mode, offset, len);
294 SYSCALL_ALIAS(sys_fallocate, SyS_fallocate);
298 * access() needs to use the real uid/gid, not the effective uid/gid.
299 * We do this by temporarily clearing all FS-related capabilities and
300 * switching the fsuid/fsgid around to the real ones.
302 SYSCALL_DEFINE3(faccessat, int, dfd, const char __user *, filename, int, mode)
304 const struct cred *old_cred;
305 struct cred *override_cred;
310 if (mode & ~S_IRWXO) /* where's F_OK, X_OK, W_OK, R_OK? */
313 override_cred = prepare_creds();
317 override_cred->fsuid = override_cred->uid;
318 override_cred->fsgid = override_cred->gid;
320 if (!issecure(SECURE_NO_SETUID_FIXUP)) {
321 /* Clear the capabilities if we switch to a non-root user */
322 if (override_cred->uid)
323 cap_clear(override_cred->cap_effective);
325 override_cred->cap_effective =
326 override_cred->cap_permitted;
329 old_cred = override_creds(override_cred);
331 res = user_path_at(dfd, filename, LOOKUP_FOLLOW, &path);
335 inode = path.dentry->d_inode;
337 if ((mode & MAY_EXEC) && S_ISREG(inode->i_mode)) {
339 * MAY_EXEC on regular files is denied if the fs is mounted
340 * with the "noexec" flag.
343 if (path.mnt->mnt_flags & MNT_NOEXEC)
344 goto out_path_release;
347 res = inode_permission(inode, mode | MAY_ACCESS);
348 /* SuS v2 requires we report a read only fs too */
349 if (res || !(mode & S_IWOTH) || special_file(inode->i_mode))
350 goto out_path_release;
352 * This is a rare case where using __mnt_is_readonly()
353 * is OK without a mnt_want/drop_write() pair. Since
354 * no actual write to the fs is performed here, we do
355 * not need to telegraph to that to anyone.
357 * By doing this, we accept that this access is
358 * inherently racy and know that the fs may change
359 * state before we even see this result.
361 if (__mnt_is_readonly(path.mnt))
367 revert_creds(old_cred);
368 put_cred(override_cred);
372 SYSCALL_DEFINE2(access, const char __user *, filename, int, mode)
374 return sys_faccessat(AT_FDCWD, filename, mode);
377 SYSCALL_DEFINE1(chdir, const char __user *, filename)
382 error = user_path_dir(filename, &path);
386 error = inode_permission(path.dentry->d_inode, MAY_EXEC | MAY_CHDIR);
390 set_fs_pwd(current->fs, &path);
398 SYSCALL_DEFINE1(fchdir, unsigned int, fd)
409 inode = file->f_path.dentry->d_inode;
412 if (!S_ISDIR(inode->i_mode))
415 error = inode_permission(inode, MAY_EXEC | MAY_CHDIR);
417 set_fs_pwd(current->fs, &file->f_path);
424 SYSCALL_DEFINE1(chroot, const char __user *, filename)
429 error = user_path_dir(filename, &path);
433 error = inode_permission(path.dentry->d_inode, MAY_EXEC | MAY_CHDIR);
438 if (!capable(CAP_SYS_CHROOT))
440 error = security_path_chroot(&path);
444 set_fs_root(current->fs, &path);
452 static int chmod_common(struct path *path, umode_t mode)
454 struct inode *inode = path->dentry->d_inode;
455 struct iattr newattrs;
458 error = mnt_want_write(path->mnt);
461 mutex_lock(&inode->i_mutex);
462 error = security_path_chmod(path, mode);
465 newattrs.ia_mode = (mode & S_IALLUGO) | (inode->i_mode & ~S_IALLUGO);
466 newattrs.ia_valid = ATTR_MODE | ATTR_CTIME;
467 error = notify_change(path->dentry, &newattrs);
469 mutex_unlock(&inode->i_mutex);
470 mnt_drop_write(path->mnt);
474 SYSCALL_DEFINE2(fchmod, unsigned int, fd, umode_t, mode)
481 audit_inode(NULL, file->f_path.dentry);
482 err = chmod_common(&file->f_path, mode);
488 SYSCALL_DEFINE3(fchmodat, int, dfd, const char __user *, filename, umode_t, mode)
493 error = user_path_at(dfd, filename, LOOKUP_FOLLOW, &path);
495 error = chmod_common(&path, mode);
501 SYSCALL_DEFINE2(chmod, const char __user *, filename, umode_t, mode)
503 return sys_fchmodat(AT_FDCWD, filename, mode);
506 static int chown_common(struct path *path, uid_t user, gid_t group)
508 struct inode *inode = path->dentry->d_inode;
510 struct iattr newattrs;
512 newattrs.ia_valid = ATTR_CTIME;
513 if (user != (uid_t) -1) {
514 newattrs.ia_valid |= ATTR_UID;
515 newattrs.ia_uid = user;
517 if (group != (gid_t) -1) {
518 newattrs.ia_valid |= ATTR_GID;
519 newattrs.ia_gid = group;
521 if (!S_ISDIR(inode->i_mode))
523 ATTR_KILL_SUID | ATTR_KILL_SGID | ATTR_KILL_PRIV;
524 mutex_lock(&inode->i_mutex);
525 error = security_path_chown(path, user, group);
527 error = notify_change(path->dentry, &newattrs);
528 mutex_unlock(&inode->i_mutex);
533 SYSCALL_DEFINE3(chown, const char __user *, filename, uid_t, user, gid_t, group)
538 error = user_path(filename, &path);
541 error = mnt_want_write(path.mnt);
544 error = chown_common(&path, user, group);
545 mnt_drop_write(path.mnt);
552 SYSCALL_DEFINE5(fchownat, int, dfd, const char __user *, filename, uid_t, user,
553 gid_t, group, int, flag)
559 if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_EMPTY_PATH)) != 0)
562 lookup_flags = (flag & AT_SYMLINK_NOFOLLOW) ? 0 : LOOKUP_FOLLOW;
563 if (flag & AT_EMPTY_PATH)
564 lookup_flags |= LOOKUP_EMPTY;
565 error = user_path_at(dfd, filename, lookup_flags, &path);
568 error = mnt_want_write(path.mnt);
571 error = chown_common(&path, user, group);
572 mnt_drop_write(path.mnt);
579 SYSCALL_DEFINE3(lchown, const char __user *, filename, uid_t, user, gid_t, group)
584 error = user_lpath(filename, &path);
587 error = mnt_want_write(path.mnt);
590 error = chown_common(&path, user, group);
591 mnt_drop_write(path.mnt);
598 SYSCALL_DEFINE3(fchown, unsigned int, fd, uid_t, user, gid_t, group)
602 struct dentry * dentry;
608 error = mnt_want_write_file(file);
611 dentry = file->f_path.dentry;
612 audit_inode(NULL, dentry);
613 error = chown_common(&file->f_path, user, group);
614 mnt_drop_write_file(file);
622 * You have to be very careful that these write
623 * counts get cleaned up in error cases and
624 * upon __fput(). This should probably never
625 * be called outside of __dentry_open().
627 static inline int __get_file_write_access(struct inode *inode,
628 struct vfsmount *mnt)
631 error = get_write_access(inode);
635 * Do not take mount writer counts on
636 * special files since no writes to
637 * the mount itself will occur.
639 if (!special_file(inode->i_mode)) {
641 * Balanced in __fput()
643 error = mnt_want_write(mnt);
645 put_write_access(inode);
650 static struct file *__dentry_open(struct dentry *dentry, struct vfsmount *mnt,
652 int (*open)(struct inode *, struct file *),
653 const struct cred *cred)
655 static const struct file_operations empty_fops = {};
659 f->f_mode = OPEN_FMODE(f->f_flags) | FMODE_LSEEK |
660 FMODE_PREAD | FMODE_PWRITE;
662 if (unlikely(f->f_flags & O_PATH))
663 f->f_mode = FMODE_PATH;
665 inode = dentry->d_inode;
666 if (f->f_mode & FMODE_WRITE) {
667 error = __get_file_write_access(inode, mnt);
670 if (!special_file(inode->i_mode))
674 f->f_mapping = inode->i_mapping;
675 f->f_path.dentry = dentry;
678 file_sb_list_add(f, inode->i_sb);
680 if (unlikely(f->f_mode & FMODE_PATH)) {
681 f->f_op = &empty_fops;
685 f->f_op = fops_get(inode->i_fop);
687 error = security_dentry_open(f, cred);
691 error = break_lease(inode, f->f_flags);
695 if (!open && f->f_op)
696 open = f->f_op->open;
698 error = open(inode, f);
702 if ((f->f_mode & (FMODE_READ | FMODE_WRITE)) == FMODE_READ)
703 i_readcount_inc(inode);
705 f->f_flags &= ~(O_CREAT | O_EXCL | O_NOCTTY | O_TRUNC);
707 file_ra_state_init(&f->f_ra, f->f_mapping->host->i_mapping);
709 /* NB: we're sure to have correct a_ops only after f_op->open */
710 if (f->f_flags & O_DIRECT) {
711 if (!f->f_mapping->a_ops ||
712 ((!f->f_mapping->a_ops->direct_IO) &&
713 (!f->f_mapping->a_ops->get_xip_mem))) {
715 f = ERR_PTR(-EINVAL);
723 if (f->f_mode & FMODE_WRITE) {
724 put_write_access(inode);
725 if (!special_file(inode->i_mode)) {
727 * We don't consider this a real
728 * mnt_want/drop_write() pair
729 * because it all happenend right
730 * here, so just reset the state.
737 f->f_path.dentry = NULL;
738 f->f_path.mnt = NULL;
743 return ERR_PTR(error);
747 * lookup_instantiate_filp - instantiates the open intent filp
748 * @nd: pointer to nameidata
749 * @dentry: pointer to dentry
750 * @open: open callback
752 * Helper for filesystems that want to use lookup open intents and pass back
753 * a fully instantiated struct file to the caller.
754 * This function is meant to be called from within a filesystem's
756 * Beware of calling it for non-regular files! Those ->open methods might block
757 * (e.g. in fifo_open), leaving you with parent locked (and in case of fifo,
758 * leading to a deadlock, as nobody can open that fifo anymore, because
759 * another process to open fifo will block on locked parent when doing lookup).
760 * Note that in case of error, nd->intent.open.file is destroyed, but the
761 * path information remains valid.
762 * If the open callback is set to NULL, then the standard f_op->open()
763 * filesystem callback is substituted.
765 struct file *lookup_instantiate_filp(struct nameidata *nd, struct dentry *dentry,
766 int (*open)(struct inode *, struct file *))
768 const struct cred *cred = current_cred();
770 if (IS_ERR(nd->intent.open.file))
774 nd->intent.open.file = __dentry_open(dget(dentry), mntget(nd->path.mnt),
775 nd->intent.open.file,
778 return nd->intent.open.file;
780 release_open_intent(nd);
781 nd->intent.open.file = ERR_CAST(dentry);
784 EXPORT_SYMBOL_GPL(lookup_instantiate_filp);
787 * nameidata_to_filp - convert a nameidata to an open filp.
788 * @nd: pointer to nameidata
791 * Note that this function destroys the original nameidata
793 struct file *nameidata_to_filp(struct nameidata *nd)
795 const struct cred *cred = current_cred();
798 /* Pick up the filp from the open intent */
799 filp = nd->intent.open.file;
800 nd->intent.open.file = NULL;
802 /* Has the filesystem initialised the file for us? */
803 if (filp->f_path.dentry == NULL) {
805 filp = __dentry_open(nd->path.dentry, nd->path.mnt, filp,
812 * dentry_open() will have done dput(dentry) and mntput(mnt) if it returns an
815 struct file *dentry_open(struct dentry *dentry, struct vfsmount *mnt, int flags,
816 const struct cred *cred)
821 validate_creds(cred);
823 /* We must always pass in a valid mount pointer. */
827 f = get_empty_filp();
831 return ERR_PTR(error);
835 return __dentry_open(dentry, mnt, f, NULL, cred);
837 EXPORT_SYMBOL(dentry_open);
839 static void __put_unused_fd(struct files_struct *files, unsigned int fd)
841 struct fdtable *fdt = files_fdtable(files);
842 __clear_open_fd(fd, fdt);
843 if (fd < files->next_fd)
847 void put_unused_fd(unsigned int fd)
849 struct files_struct *files = current->files;
850 spin_lock(&files->file_lock);
851 __put_unused_fd(files, fd);
852 spin_unlock(&files->file_lock);
855 EXPORT_SYMBOL(put_unused_fd);
858 * Install a file pointer in the fd array.
860 * The VFS is full of places where we drop the files lock between
861 * setting the open_fds bitmap and installing the file in the file
862 * array. At any such point, we are vulnerable to a dup2() race
863 * installing a file in the array before us. We need to detect this and
864 * fput() the struct file we are about to overwrite in this case.
866 * It should never happen - if we allow dup2() do it, _really_ bad things
870 void fd_install(unsigned int fd, struct file *file)
872 struct files_struct *files = current->files;
874 spin_lock(&files->file_lock);
875 fdt = files_fdtable(files);
876 BUG_ON(fdt->fd[fd] != NULL);
877 rcu_assign_pointer(fdt->fd[fd], file);
878 spin_unlock(&files->file_lock);
881 EXPORT_SYMBOL(fd_install);
883 static inline int build_open_flags(int flags, umode_t mode, struct open_flags *op)
885 int lookup_flags = 0;
888 if (!(flags & O_CREAT))
892 /* Must never be set by userspace */
893 flags &= ~FMODE_NONOTIFY;
896 * O_SYNC is implemented as __O_SYNC|O_DSYNC. As many places only
897 * check for O_DSYNC if the need any syncing at all we enforce it's
898 * always set instead of having to deal with possibly weird behaviour
899 * for malicious applications setting only __O_SYNC.
901 if (flags & __O_SYNC)
905 * If we have O_PATH in the open flag. Then we
906 * cannot have anything other than the below set of flags
908 if (flags & O_PATH) {
909 flags &= O_DIRECTORY | O_NOFOLLOW | O_PATH;
912 acc_mode = MAY_OPEN | ACC_MODE(flags);
915 op->open_flag = flags;
917 /* O_TRUNC implies we need access checks for write permissions */
919 acc_mode |= MAY_WRITE;
921 /* Allow the LSM permission hook to distinguish append
922 access from general write access. */
923 if (flags & O_APPEND)
924 acc_mode |= MAY_APPEND;
926 op->acc_mode = acc_mode;
928 op->intent = flags & O_PATH ? 0 : LOOKUP_OPEN;
930 if (flags & O_CREAT) {
931 op->intent |= LOOKUP_CREATE;
933 op->intent |= LOOKUP_EXCL;
936 if (flags & O_DIRECTORY)
937 lookup_flags |= LOOKUP_DIRECTORY;
938 if (!(flags & O_NOFOLLOW))
939 lookup_flags |= LOOKUP_FOLLOW;
944 * filp_open - open file and return file pointer
946 * @filename: path to open
947 * @flags: open flags as per the open(2) second argument
948 * @mode: mode for the new file if O_CREAT is set, else ignored
950 * This is the helper to open a file from kernelspace if you really
951 * have to. But in generally you should not do this, so please move
952 * along, nothing to see here..
954 struct file *filp_open(const char *filename, int flags, umode_t mode)
956 struct open_flags op;
957 int lookup = build_open_flags(flags, mode, &op);
958 return do_filp_open(AT_FDCWD, filename, &op, lookup);
960 EXPORT_SYMBOL(filp_open);
962 struct file *file_open_root(struct dentry *dentry, struct vfsmount *mnt,
963 const char *filename, int flags)
965 struct open_flags op;
966 int lookup = build_open_flags(flags, 0, &op);
968 return ERR_PTR(-EINVAL);
969 if (!filename && (flags & O_DIRECTORY))
970 if (!dentry->d_inode->i_op->lookup)
971 return ERR_PTR(-ENOTDIR);
972 return do_file_open_root(dentry, mnt, filename, &op, lookup);
974 EXPORT_SYMBOL(file_open_root);
976 long do_sys_open(int dfd, const char __user *filename, int flags, umode_t mode)
978 struct open_flags op;
979 int lookup = build_open_flags(flags, mode, &op);
980 char *tmp = getname(filename);
981 int fd = PTR_ERR(tmp);
984 fd = get_unused_fd_flags(flags);
986 struct file *f = do_filp_open(dfd, tmp, &op, lookup);
993 trace_do_sys_open(tmp, flags, mode);
1001 SYSCALL_DEFINE3(open, const char __user *, filename, int, flags, umode_t, mode)
1005 if (force_o_largefile())
1006 flags |= O_LARGEFILE;
1008 ret = do_sys_open(AT_FDCWD, filename, flags, mode);
1009 /* avoid REGPARM breakage on x86: */
1010 asmlinkage_protect(3, ret, filename, flags, mode);
1014 SYSCALL_DEFINE4(openat, int, dfd, const char __user *, filename, int, flags,
1019 if (force_o_largefile())
1020 flags |= O_LARGEFILE;
1022 ret = do_sys_open(dfd, filename, flags, mode);
1023 /* avoid REGPARM breakage on x86: */
1024 asmlinkage_protect(4, ret, dfd, filename, flags, mode);
1031 * For backward compatibility? Maybe this should be moved
1032 * into arch/i386 instead?
1034 SYSCALL_DEFINE2(creat, const char __user *, pathname, umode_t, mode)
1036 return sys_open(pathname, O_CREAT | O_WRONLY | O_TRUNC, mode);
1042 * "id" is the POSIX thread ID. We use the
1043 * files pointer for this..
1045 int filp_close(struct file *filp, fl_owner_t id)
1049 if (!file_count(filp)) {
1050 printk(KERN_ERR "VFS: Close: file count is 0\n");
1054 if (filp->f_op && filp->f_op->flush)
1055 retval = filp->f_op->flush(filp, id);
1057 if (likely(!(filp->f_mode & FMODE_PATH))) {
1058 dnotify_flush(filp, id);
1059 locks_remove_posix(filp, id);
1065 EXPORT_SYMBOL(filp_close);
1068 * Careful here! We test whether the file pointer is NULL before
1069 * releasing the fd. This ensures that one clone task can't release
1070 * an fd while another clone is opening it.
1072 SYSCALL_DEFINE1(close, unsigned int, fd)
1075 struct files_struct *files = current->files;
1076 struct fdtable *fdt;
1079 spin_lock(&files->file_lock);
1080 fdt = files_fdtable(files);
1081 if (fd >= fdt->max_fds)
1086 rcu_assign_pointer(fdt->fd[fd], NULL);
1087 __clear_close_on_exec(fd, fdt);
1088 __put_unused_fd(files, fd);
1089 spin_unlock(&files->file_lock);
1090 retval = filp_close(filp, files);
1092 /* can't restart close syscall because file table entry was cleared */
1093 if (unlikely(retval == -ERESTARTSYS ||
1094 retval == -ERESTARTNOINTR ||
1095 retval == -ERESTARTNOHAND ||
1096 retval == -ERESTART_RESTARTBLOCK))
1102 spin_unlock(&files->file_lock);
1105 EXPORT_SYMBOL(sys_close);
1108 * This routine simulates a hangup on the tty, to arrange that users
1109 * are given clean terminals at login time.
1111 SYSCALL_DEFINE0(vhangup)
1113 if (capable(CAP_SYS_TTY_CONFIG)) {
1121 * Called when an inode is about to be open.
1122 * We use this to disallow opening large files on 32bit systems if
1123 * the caller didn't specify O_LARGEFILE. On 64bit systems we force
1124 * on this flag in sys_open.
1126 int generic_file_open(struct inode * inode, struct file * filp)
1128 if (!(filp->f_flags & O_LARGEFILE) && i_size_read(inode) > MAX_NON_LFS)
1133 EXPORT_SYMBOL(generic_file_open);
1136 * This is used by subsystems that don't want seekable
1137 * file descriptors. The function is not supposed to ever fail, the only
1138 * reason it returns an 'int' and not 'void' is so that it can be plugged
1139 * directly into file_operations structure.
1141 int nonseekable_open(struct inode *inode, struct file *filp)
1143 filp->f_mode &= ~(FMODE_LSEEK | FMODE_PREAD | FMODE_PWRITE);
1147 EXPORT_SYMBOL(nonseekable_open);